Gene expression profiling measures the relative abundance of a large number of individual mRNA species within the context of a total mRNA population that has been isolated and purified from a target biological sample. The technology for achieving this evolved from low-Throughput membrane-based probeto- target hybridization assays (northern and Southern blotting) commonly used in molecular biology labs since the 1980s (reviewed in 1). Scaling these assays down in size and up in number by redesigning the assay into a " targetto- probe" hybridization has enabled efficient and highly parallel genomewide gene expression analysis. The technology platforms that emerged are generically called microarrays. A comprehensive review of microarrays, their history, application, and analysis is given by Schena (2). The goal of gene expression profiling, achieved by microarray analysis, is to aid the biologist in identifying groups of genes that are functionally associated with certain biological processes. "Guilt by association" is a classical approach that is based on correlation analysis of mRNA abundance measures taken from biological samples subjected to various experimental factors (variables or conditions) (3). Typical variables are genotype (e.g., mutant vs. wild type), tissue, time, and treatment, or combinations of these. Visual representation of the results of correlation analysis are frequently dendrograms and "heat-maps" of expression values ordered (clustered) for each gene and/or each condition based on statistical measures of similarity. The underlying data provides the leads for hypothesis development and further experimentation. In this chapter we have chosen to restrict our discussion to the two current and most commonly used microarray formats: spotted arrays and Affymetrix GeneChips. The basic principles of microarray analysis are outlined in Fig. 14.1A, B. As many components of a microarray experiment - e.g., sample choice, preparation and analysis-Are largely similar across both of these platforms, we consider these aspects together, but discuss the different technologies separately. We conclude by considering examples and summarize the need for adequate data standards and public access.